Size charts for automatic size recommendation: how to prepare them (CSV, categories, shoes)

Cabina does not know what “M” means, and it does not need to. When it recommends a size it compares the shopper’s measurements — height, bust, waist, hips, estimated from the photo or typed in by hand — with the ranges in centimetres you declared for each row of the chart, and picks the row they fall into. If the measurements land between two rows, it says so: “Between M and L — we recommend L”, with a confidence percentage.

Hence the rule that governs everything else: the label is free, the range is the substance. “34”, “40”, “age 8”, “XL” are just names; the order of sizes comes from the centimetres, smallest to largest. There is no European or American scale hard-wired anywhere: your chart is the only truth, and that is why it deserves care.

A chart is a CSV file with nine columns, in this order: size, height_min, height_max, bust_min, bust_max, waist_min, waist_max, hips_min, hips_max. Body measurements are whole centimetres. One row per size, for example: S,160,168,84,92,66,72,90,96.

• A column you do not need stays at 0. For trousers, height and bust may not matter: put 0 and 0, and that dimension is ignored in the comparison.
• The size name is text, up to 20 characters: numbers, letters, “age 4”, “38½”.
• Ranges must make sense: the minimum cannot exceed the maximum, and rows that overlap too much make the recommendation uncertain.

You upload the file from the dashboard, where you also find manual entry row by row. If you prefer automation, the same charts are available through the REST API and the MCP server.

The point most often missed: a top and a pair of trousers are not measured the same way, and in Europe they are not even named the same way. A top is S, M, L; trousers are 40, 42, 44; a shoe is 38, 39, 40. With a single chart, a shopper trying on trousers gets “M” — a label that does not exist in the product page’s selector.

That is why every chart declares which garment it applies to: tops, trousers and skirts, dresses, outerwear, shoes — or any garment. Cabina recognises the garment category on the page by itself (it already does so to generate the try-on) and uses the right chart. Outerwear without a chart of its own falls back to the tops chart; when nothing specific exists, the generic chart is used.

In practice: a chart with numeric sizes for trousers, with waist and hips declared, and the recommendation will read “42” — exactly what the shopper can select.

If you also sell for kids, you do not need a separate store: upload two charts of the same category — adult trousers 40–52 with heights 160–195, kids’ trousers “age 4” to “age 14” with heights 100–160. Among several candidate charts, the person’s measurements decide, not the upload order: someone 130 cm tall lands in the second, someone 175 cm tall in the first.

The same goes for two brands with different fits: two charts, and the one the measurements fit best wins.

For footwear, height and bust say nothing. The shoe chart uses two extra, optional columns: foot_min and foot_max, the foot length in centimetres, decimals allowed. A row looks like: 42,0,0,0,0,0,0,0,0,26.5,27.1 — every body measurement at 0, only the foot declared.

When the item on the page is a shoe, the fitting room asks the shopper for their foot length; if they leave it blank, it is estimated from height and treated as an estimated measurement, with lower confidence. Measuring takes a minute: foot on the floor against a wall, and a ruler from the wall to the tip of the big toe.

The result shows the recommended size with its confidence, and — where the page has a recognisable selector — that size is preselected: the “Add to cart” after the try-on starts with the right variant, without the shopper reapplying it by hand. If the chart does not cover their measurements, no recommendation appears: no advice beats wrong advice.

The recommendation depends on measurements the shopper can correct, which is why the measurements form stays within reach. How the recommendation cuts returns is the subject of Reducing returns: the wrong size and how to avoid it.

• Nine columns in the right order, whole centimetres (decimals only for the foot).
• One chart per category: tops, trousers, dresses, outerwear, shoes.
• Trousers and shoes with numeric labels, as in the product page’s selector.
• Dimensions that do not matter at 0, not made up.
• Adults and kids in two charts of the same category.
• A test on the free plan with one item per category, to see that the recommendation appears and is selectable.